1. Field of the Invention
The present invention relates to data communications systems and, in particular, to auto-threshold/adaptive equalizer circuitry for an integrated services digital network (ISDN) S-interface.
2. Discussion of the Prior Art
Although data communications systems are designed to yield zero intersymbol interference (ISI) during transmission, some amount of residual ISI inevitably occurs due to imperfections in the design or changing channel characteristics. Typically, adjustable filters are included within the system to compensate for this distortion. The process of correcting channel induced distortion is called equalization.
The most commonly used form of adjustable equalizer has been the transversal filter. A transversal filter consists, essentially, of a delay line which is tapped at predefined intervals. Each tap is connected through a variable gain device to a summing amplifier. By specifying the output value of the equalizer, the required gain settings can be determined. This type of equalizer is called a zero forcing equalizer since the equalizer output has an equal number of zero values on either size. Zero forcing equalizers are optimum in that they minimize peak ISI. The primary problem with zero forcing equalizers is that they increase the noise power at the input of the A/D converter typically included in the receiver system.
The design and adjustments of the tap gains of the zero forcing equalizer described above involve the solution of a set of simultaneous equations. In the manual made, the "trimming" of the equalizer requires that a test pulse be sent through the system, the receiving filter output be measured at appropriate sampling times, the tap gains be solved for using appropriate mathematical relationships and, finally, that the gain on each tap be set manually.
To eliminate the need for these time-consuming manual adjustments, automatic systems for setting the tap gains have been developed. These systems are usually divided into two groups. Preset equalizer systems use a special sequence of pulses prior to or during breaks in data transmission. Adaptive equalizers use iterative techniques to arrive at optimum tap gains.
Preset equalizers require an iterative "training" procedure, sometimes involving hundreds of test pulses, to arrive at the desired tap gains. A major problem in "training" a preset equalizer is the presence of noise in the observed equalizer output values. While the effects of noise can be somewhat minimized by averaging the measured output values over a number of test pulses, this averaging technique slows down the rate of convergence.
In an adaptive equalizer, tap gain corrections are continually estimated during the normal course of data transmission. Thus, adaptive equalizers have the ability to adapt to changes during data transmission and eliminate the need for the long training procedures required with preset equalizers. Adaptive equalizers are also more accurate, versatile and cheaper than preset equalizers.
Theoretically, in a simple adaptive equalizer, the output of the equalizer at sampling times should be either plus or minus a preselected value: plus, if the actual input bit corresponding to the sampling time is 1; minus, if the input bit is 0. In an actual system, the equalizer output values will vary about the preselected value, due to ISI, depending on the input sequence. If the ISI is not very large, the transmitted data can be decoded and a sequence of ideal or desired output values is generated. An estimate of the error sequence required for adjusting the tap gains is then generated based on a comparison of the ideal outputs with the actual measured outputs.